First, the "PtByPt" means that it works on one datapoint at a time. The data is builtup internally into an array that will be processed--sort of like the running average example but with an impulse response operation instead of a simple mean.
Second, the stimulus is the signal going into the network for which you are trying to measure the impluse response. The response signal is the data coming out of the network. The basic idea is that the algorithm calculates the networks impulse response by comparing what went into it with what came out of it.
The only difference I might make is that it sounds like you already have all your data in array form. If that is the case, you might want to make use of the non-PtByPt version of the function.
Mike...